How to organize your thoughts before speaking

When someone asks you a question in a meeting, you probably start talking to figure out what you think. The people listening get the whole process: false starts, background that goes nowhere, and the actual point somewhere around sentence four.

This article covers a ten-second routine for deciding your point, and one structure that handles most workplace situations. It also covers how to signal that structure while you talk, and a method for rehearsing ideas that matter. Use the ten-second routine in your next meeting; save the rehearsal method for updates and talks with real stakes.

The advice comes from Matt Abrahams, who has taught strategic communication at Stanford's Graduate School of Business for 25 years; Wes Kao, co-founder of Maven and altMBA, who coaches executives on communication; and Tristan de Montebello, co-founder of the speaking school Ultraspeaking.

Decide the answer before the background

You don't need a full script before you speak. You need the one sentence that answers the question, plus whatever supports it: two reasons, one example, or a decision with its tradeoff.

Before you open your mouth, silently finish this sentence: "The main thing I want them to know is..." If you can't finish it, ask a clarifying question or ask for a moment. Once you know the point, choosing which background to include gets much easier, because background either supports the point or it doesn't belong.

When your manager asks "Are we on track?", that routine produces an opening like: "We're on track for the date, with one unresolved legal approval." The timeline and the history can follow if anyone wants them.

A structure does some of the thinking for you

Abrahams argues that structure matters most exactly when you have no time to prepare. Under pressure, most people dump information in the order they think of it.

His point in the same stretch of the conversation: our brains handle connected ideas much better than lists. A structure is just a beginning, a middle, and an end that belong together. He compares it to a recipe: "If I have a recipe and I have good ingredients, I can cook a good meal." The structure decides how to say it, so the only decision left while you're talking is what to put into it. That helps the listener too, because they know what each sentence is doing.

Use what, so what, now what

Abrahams teaches several structures (PREP, problem-solution-benefit), but the one he recommends most broadly is a three-question shape:

• What: the fact, change, or observation.
• So what: why it matters to the people you're talking to.
• Now what: the next action, recommendation, or question.

Here's the shape carrying a product risk: "Trial conversion fell from 18 to 13 percent this month. That puts the quarter's new-revenue target at risk even if traffic holds. I recommend we pause the new acquisition test and spend this week diagnosing activation." Three sentences, and the listener knows the fact, the stakes, and the ask.

Match the shape to the job

What, so what, now what covers most situations, but some conversations have a more specific job. A few variants worth having ready:

Don't skip the tradeoff in a recommendation. Saying it shows you've already thought about the objection someone was about to raise, which shortens the discussion.

Say the structure out loud

Nobody can see your outline. Wes Kao's fix is signposting: specific words that tell the listener what the next sentence is doing.

She's describing writing there, but she says in the same answer that it works verbally too: in a demo, "the most important part to pay attention to is..." pulls back anyone whose attention drifted. Numbering helps the same way: "There are three reasons" tells people how long to keep listening.

Use signposts only when they carry information. Saying "firstly, secondly, thirdly" around trivial items makes a short answer feel like a speech. A good signpost tells people whether they're about to hear evidence, contrast, consequence, or action.

Rehearse important ideas by speaking them

A written paragraph hides sentences that are hard to say out loud. For ideas you'll deliver more than once, a big update or a talk, de Montebello replaces writing and memorizing with something he calls the Accordion Method. His argument against scripts: memorization is a chain, and losing one link leaves you stuck.

The version he walks through in the episode, with a timer and a rule that you talk to the end each time:

1. Speak for three minutes. Get every idea out, however rough.
2. Repeat in two minutes. Cut the tangents and anything that didn't feel right.
3. Repeat in one minute. Keep the central point and the strongest proof.
4. Repeat in 30 seconds. This is the version you'd give an executive in a hallway.
5. Expand again. Add back only the detail you now know you want.

By the 30-second round you know exactly which pieces are essential, because you've had to choose them four times. When you expand again, you add detail without burying them.

When the question surprises you

The routine for a question you didn't see coming is short: pause, decide what kind of answer is being asked for, clarify if the question is ambiguous, then state the point you can actually support and add one example. A clarifying question is faster than a polished answer to the wrong problem: "When you say adoption, are you asking how many teams started, or how many are still active after a month?"

If you don't have the full answer, say what you do have and bound the rest: "I don't have the retention number in front of me. Weekly active use has risen for four weeks. I'll send the exact retention figure after the meeting."

Answering on the spot, knowing when to stop, and lines that buy you time are covered in depth in how to stop rambling in meetings, including Abrahams' answer-example-relevance structure from the same episode.

Frequently asked questions

How do I stop overexplaining?

Lead with the conclusion, give one example, and say why it matters. Then offer more detail instead of including it: "I can walk through the numbers if useful." Letting the group opt in removes the urge to front-load everything.

What if I think better out loud?

Do the out-loud thinking before the meeting, in a voice note or with a colleague. If you genuinely need to explore in the meeting, label it: "I'm thinking this through, so treat this as a hypothesis." That tells people not to file it as your conclusion.

How many points should I make at once?

One main point with two or three supporting items is usually enough. If the topic honestly needs more, group the details into a few named categories so listeners have somewhere to put them.

How do I remember a structure under pressure?

Practice one until it's your default. What, so what, now what covers updates, recommendations, and explanations well enough that you rarely need another. Use it in low-stakes standups first so it's available when the stakes go up.

The one-page routine